reviewing-translations
Warn
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script validate_review_artifact.py employs importlib.util to dynamically load invariants.py and an external policy script from paths computed at runtime. This involves dynamic loading from computed paths, including paths traversing outside the skill's immediate directory.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted translation data, creating a potential attack surface for embedded malicious instructions. Evidence Chain: 1) Ingestion points: source and target translation artifacts as described in SKILL.md. 2) Boundary markers: implementation of strict JSON output schema requirements and delivery boundaries. 3) Capability inventory: local script execution for validation using python3. 4) Sanitization: automated structural and invariant validation performed by internal scripts.
- [COMMAND_EXECUTION]: The skill's instructions direct the agent to execute a local Python validator (validate_review_artifact.py) and provide specific command-line arguments that include user-supplied file paths.
Audit Metadata